Mapping and inspecting natural small bodies, such as asteroids and comets, is essential for planetary science and target characterization. Unlike artificial targets, small bodies often present highly irregular gravity fields and poorly characterized environments. These features make it significantly challenging to design trajectories that optimize surface coverage while maintaining spacecraft safety and respecting operational constraints.

Much of the existing literature on autonomous trajectory planning for mapping has been developed with artificial targets in mind, driven by the interest in debris-removal missions and inspection of unknown objects in Earth orbit. Such methods assume well-modeled dynamical environments and stable relative motion and, therefore, do not directly generalize to the highly perturbed dynamical environment surrounding natural small bodies. Among these approaches, reinforcement-learning [1] offers adaptability but lacks systematic constraint-handling mechanisms, does not exploit available knowledge of a world model, and must be retrained for each new environment. Waypoint-based approaches, while simple to implement, often become infeasible in this context, as predetermined observation points may not correspond to achievable trajectories when the gravitational field is not negligible and highly irregular. Finally, heuristic global-optimization methods and sampling-based planners [2] can identify feasible solutions but often exceed the limited computational resources available on flight hardware.

These limitations motivate a deterministic, model-based framework capable of optimizing inspection trajectories while efficiently handling nonlinear dynamics. In line with ongoing research on autonomous guidance with convex formulations for onboard Model Predictive Control [3], this work, carried out as part of the research activities of the ASTRA scientific laboratory, advances a convex-optimization-based approach tailored to small-body mapping scenarios.

Due to the nonlinear nature of  both dynamics and collision avoidance constraints, the proposed framework employs Sequential Convex Programming (SCP) using a trust-region solver with convergence guarantees [4]. This approach incorporates the affine linearization of the spacecraft dynamics according to an onboard model of the irregular gravity field while efficiently minimizing a nonconvex, continuously differentiable mapping objective. A relevant contribution is the reformulation of the mapping cost to avoid reliance on predefined observation waypoints. Instead, the trajectory is optimized by penalizing the distance to imaging-related constraints associated with a selected set of surface landmarks to observe on the asteroid. These include a minimum and maximum range for correct observation, a maximum off-nadir angle, and correct illumination of the scene. A multiplicative, differentiable cost encourages sequential proximity to any of the feasible imaging regions, while ensuring the computation of dynamically feasible trajectories in irregular gravitational fields.

The second contribution is the extension of this formulation to distributed multi-agent scenarios, relevant for cooperative small-body inspection campaigns. Each spacecraft solves a local SCP subproblem based on its own dynamics while coordinating shared geometric variables, such as expected landmark coverage or inter-spacecraft distances, through consensus-based exchanges. This goal-coupled formulation enables cooperation between the agents to achieve efficient and safe collaborative mapping.

The framework is then tested in high-fidelity simulation environments, validating the improved mapping performance relative to typical propellant-free trajectory designs for small-body exploration and demonstrating how the coupling between objectives and constraints is handled within the distributed implementation. The simulations highlight improvements in both revisit frequency of selected surface points and the overall percentage of mapped landmarks. These results are obtained for a set of past mission targets spanning a wide range of sizes and rotational parameters, confirming the robustness of the method to different dynamical conditions.

[1] Brandonisio, A., Capra, L., and Lavagna, M., “Deep reinforcement learning spacecraft guidance with state uncertainty for autonomous shape reconstruction of uncooperative target,” Advances in Space Research, Vol. 73, No. 11, 2024, pp. 5741–5755. https://doi.org/10.1016/j.asr.2023.07.007.

[2] Capolupo, F., Simeon, T., and Berges, J.-C., “Heuristic Guidance Techniques for the Exploration of Small Celestial Bodies,” IFAC-PapersOnLine, Vol. 50, No. 1, 2017, pp. 8279–8284. https://doi.org/https://doi.org/10.1016/j.ifacol.2017.08.1401, 20th IFAC World Congress.

[3] Rizza, A., D’Amico, S., and Topputo, F., “Goal-Oriented Trajectory Refinement for Asteroid Mapping Using Sequential Convex Programming,” Journal of Guidance, Control, and Dynamics, Vol. 0, No. 0, 2025, pp. 1–15. https://doi.org/10.2514/1.G008976.

[4] Malyuta, D., Reynolds, T. P., Szmuk, M., Lew, T., Bonalli, R., Pavone, M., and Aç?kme?e, B., “Convex Optimization for Trajectory Generation: A Tutorial on Generating Dynamically Feasible Trajectories Reliably and Efficiently,” IEEE Control Systems Magazine, Vol. 42, No. 5, 2022, pp. 40–113. https://doi.org/10.1109/MCS.2022.3187542.